Hi Sal,
2012/8/7 sal <svfe...@gmail.com>:
> According to benchmarks, ie:
http://attractivechaos.github.com/plb/
> Shedskin is still 2-4x slower than 'real' C and C++. (Which is still near
> top of that list, so very impressive).
> Any ideas as to what the bottlenecks are currently?
This has been discussed previously. See
https://groups.google.com/d/topic/shedskin-discuss/CAdTn5EL3F8/discus...
.
> I wanted to ping the
> list before starting up my profiler and trying to recreate those test cases.
> (Side note I'm a C++ coder, optimization is fun so I might try my hand at
> some of it.)
There are some known possible improvements to the GC behaviour as well
as suboptimal implementations of some builtins (immutable strings,
slices, files…) and some issues related to the design which may be
less easy to solve (heap-allocated scalars, numerous indirections…). I
don't remember what were the bottlenecks in this case nor if the
benchmark still reflects the current state of shedskin.
Best regards,
--
Jérémie
